KR102101409B1 - Method of extending ontology data based on topic modeling - Google Patents

Abstract

According to the present invention, words are classified for each topic by using a topic modelling method and when a user forms key words and layer information in a specific topic, parse tree information is generated to extract words similar to the layer information of the key words formed by the user. In addition, word vector information (word embedding) is extracted by using the extracted words and words with high connectivity are extracted by using similarity between the key words selected by the user and the word vector information. The series of processes are repeated to extend ontology (sematic-graph) information having a layer structure and connectivity information for words of interest of the user. A method of extracting the word vector information can be changed to a parse tree method, wherein the ontology information selected by the user can be extended by using layer information of words extracted by parse analysis and applying an SVD method.

Description

Method of extending ontology data based on topic modeling method {Method of extending ontology data based on topic modeling}

The present invention relates to a method of extending ontology information (ontology, sematic graph) representing hierarchical structure and connection information of a key word selected by a user by using a topic modeling method in the field of natural language processing and image processing analysis to be.

Feature vectors can be extracted from text and image information to classify each information by topic. For this topic modeling, the SVD (singular value decomposition) and LDA (latent dirichlet allocation) methods can be applied.

Ontology information may be configured by using linkage and hierarchy between each major information using text and image information classified as topic modeling.

In the US9892194 patent, a method of determining the hierarchical structure by applying a weight based on an important paragraph was proposed, and in the US9449051 patent, a hierarchical structure was determined by taking into account the frequency of words by topic, and the US10216829 patent calculated the topic probability for a group of words Did.

I would like to present a method of extending ontology (knowledge-graph, sematic-graph) information that satisfies hierarchy and has high relation to users' words of interest in big data documents.

By using big data document information as topic modeling, words can be classified by topic. When the user constructs key words and hierarchical information on a specific topic, the system creates parse tree information and extracts words similar to the hierarchical information of the key words configured by the user. Using the extracted word, the word vector information (word embedding) is extracted, and a word having high connectivity is extracted using the similarity between the key words selected by the user and the word vector information. By repeating this series of processes, ontology (sematic-graph) information having hierarchical structure and connectivity information is expanded for a user's word of interest. The method of extracting the word vector information may be changed by a parse tree method. In this case, the ontology information selected by the user may be extended by using hierarchical information of words extracted by parsing.

According to the user's main interest word, the importance of the word is determined relatively, and by reflecting the hierarchical information and connectivity information between words defined by the user, the word of interest, hierarchical information, and a word with high connectivity are effectively expanded, thereby It is possible to generate ontology information.

1 is a conceptual diagram for topic modeling of the present invention.
2 shows an SVD matrix according to an embodiment of the present invention.
3 illustrates a method of extracting a word having high similarity and expanding ontology information by using word vector information by reflecting hierarchical information of a user interest word according to an embodiment of the present invention.
4 illustrates hierarchical information of a user interest word according to an embodiment of the present invention. It shows how to extend ontology information by extracting words with high connection of hierarchical information by applying SVD method to syntax tree information.
5 is an example of a syntax tree according to an embodiment of the present invention.

Terms used in the present specification will be briefly described, and the present invention will be described in detail. The terminology used in the present invention was selected from the general terms that are currently widely used while considering the functions in the present invention, but this may vary according to the intention or precedent of a person skilled in the art or the appearance of new technologies. In addition, in certain cases, some terms are arbitrarily selected by the applicant, and in this case, their meanings will be described in detail in the description of the corresponding invention. Therefore, the terms used in the present invention should be defined based on the meanings of the terms and the contents of the present invention, not simply the names of the terms.

When a part of the specification "includes" a certain component, this means that other components may be further included instead of excluding other components unless specifically stated otherwise. In addition, terms such as “... unit” and “module” described in the specification mean a unit that processes at least one function or operation, which may be implemented in hardware or software, or a combination of hardware and software. .

Hereinafter, embodiments of the present invention will be described in detail with reference to the accompanying drawings so that those skilled in the art to which the present invention pertains may easily practice. However, the present invention can be implemented in many different forms and is not limited to the embodiments described herein. In addition, in order to clearly describe the present invention in the drawings, parts irrelevant to the description are omitted, and like reference numerals are assigned to similar parts throughout the specification.

Hereinafter, the present invention will be described in detail with reference to the accompanying drawings.

1 is a conceptual diagram for topic modeling of the present invention. All words in each document are categorized according to user-defined topics. Therefore, words related to various topics may be distributed in one document.

Singular Value Decomposition (SVD) or Latent Dirichlet Allocation (LDA) is used as a topic classification method. SVD, also known as singular value decomposition, is one of the methods of matrix decomposition. In addition to lowering the computational cost, it is possible to check the deep meaning that was not revealed in existing data by deleting and compressing data with low information. 2 shows an SVD matrix according to an embodiment of the present invention. The equation below is the SVD determinant.

U and V are orthogonal matrices and S is a diagonal diagonal matrix.

5 is an example of a syntax tree according to an embodiment of the present invention. The syntax tree analyzes the morphemes of sentences through syntax analysis, and analyzes them in the form of word parts of speech and sentence components. Therefore, hierarchical information and connectivity of words in a sentence can be analyzed using information of a syntax tree.

3 illustrates a method of extracting a word having high similarity and expanding ontology information by using word vector information by reflecting hierarchical information of a user interest word according to an embodiment of the present invention.

In step S310 of FIG. 3, the number of topic models is determined for topic modeling.

In step S320, a topic extracted by a topic modeling method is selected, and words included in the topic are selected. In this case, words to be used may be selected using importance information for each word or probability information to be included in a topic for each word. Therefore, the flowchart of FIG. 3 is performed for each topic.

Step S330 is a process in which the user selects a word of interest. At this time, the user can select a word or a group of words as a word of interest. In step S340, the user defines hierarchical information (syntax tree 1) and connection information of the key word of interest. Connection information can be expressed by various semantic connection relationships between two words.

In step S340, hierarchical information and connection information between a key word of interest and a corresponding word may be defined differently for each user. Therefore, it is a step of setting the search direction of the system by defining the word of interest and ontology information of the user.

In step S350, the syntax tree-B having hierarchical information similar to the syntax tree-A defined in step S340 is extracted by using sentences containing a user interest word. At this time, in step S360, a valid word is determined using the similarity of the hierarchical information of the syntax tree-A and the syntax tree-B and word distance information of the word of interest to the user on the syntax tree.

Step S370 is a step of generating word vector information (word2vec, word-embedding) using words of the extracted valid syntax information. The word vector information represents similarity information of all words with a value between -1 and 1, or between 0 and 1. The higher the similarity value, the higher the similarity between the two words. Using the word vector information extracted in step S370, a set of words having high similarity to the word of interest to the user may be extracted. At this time, valid word sets can be determined by a statistical processing method. N words with high similarity to each user's interest word are selected and extracted as a group of valid similar words. The number of effective pseudoword sets is determined by the number of words of interest. Statistical values such as mean-A, standard deviation-A, and variance-A of each effective pseudoword set are extracted, and the information of each effective similar word and the mean-A is used, so that the sum of deviations of each effective similar word is small A group of valid similar words can be scored in order. In addition, the effective similar word group may be adjusted so that the variation of the variance value between each effective similar word group has a value lower than a certain criterion. Therefore, the variance value and the variance of the variance value can be used as a user set value that can determine an effective similar word . The user setting value may be changed and applied each time step S370 is performed. In addition, when a plurality of variance values of the group of effective similar words exist, it is possible to determine whether a certain significance probability is satisfied through an ANOVA method.

Step S380 is a step of including the effective similar word selected in S370 into an ontology composed of user interest words. When the step S380 is completed, the step S330 is performed again based on the extended ontology information. If the step S330 is repeated again, the step S380 may include a step for the user to determine the effective similar word to be included in the user ontology, or when the statistical value of the valid similar word satisfies the threshold, it is included in the user ontology. , It can be set to automatically perform step S330 repeatedly. In addition, when the step S340 is performed again, the updated user ontology information may be used, or the user may modify the ontology information. Therefore, when steps S330-S380 are repeatedly performed, the ontology information defined by the user is continuously updated, and the valid similar words are updated with word vector information. In addition, the steps S330-S380 may be repeatedly performed until no more similar words are extracted.

In FIG. 3, step S360 may branch to step G1. The condition of branching is a case where the ontology expansion word is determined by using word hierarchical information of the syntax tree determined in step S360, in addition to a method having high similarity to a user interest word as word vector information.

Step S410 is a step of expressing the effective word existing within a certain distance from the user interest word in the hierarchical information of the syntax tree as matrix information. The matrix value is configured in the form of the inverse of the distance information of the effective word. The shorter the distance between the user's interest word and the effective word in the syntax tree, the higher the connectivity, so the smaller the distance information, the more effective.

Before constructing a valid matrix, a network analysis method can be applied by using link information existing between valid words in a document. The effective word can be selectively extracted based on the link score extracted from the network information. Therefore, in step S410, in order to extract the user effective word, the distance between the user's interest word and the effective word candidate and the link score or rank of network information may be performed by referring to the user setting value. In addition, the user setting value may be changed and applied every time step S360 or step S410 is performed.

In step S420, in step S410, all generated matrix information is grouped into an entire matrix, and an SVD method is applied thereto.

In step S430, it is a step of selecting the selected SVD (truncated SVD) using only the upper value of the eigenvalues value obtained through the SVD in step S420. At this time, in step S440, the effective matrix value distribution and diagonal matrix information of the selected SVD may be separated into a plurality of subgroups. In addition, step S440 is a step of extracting related word (entity) information from the hierarchical information of the selected SVD including a plurality of subgroups determined in step S430.

Step S450 is a step of combining ontology information composed of user interest words and a selectable SVD subgroup having high connectivity. It can be combined with user ontology information using key words and hierarchical information included in the selected SVD subgroup. At this time, the selected SVD subgroup may be combined using distance information between the closest user's interest words, or the user may designate a word of interest to which the selected SVD subgroup will be combined. In step S450, by combining a selected SVD subgroup consisting of one or more words with a user's word of interest, the user can more clearly understand the meaning of the recommended word group, and can easily determine expansion of user ontology information. When step S450 is performed, step S330 of FIG. 3 is re-executed using the extended user ontology information. Therefore, as steps S330 to S450 are repeatedly performed, when the user ontology information is continuously expanded and a new valid word candidate is no longer generated, it is terminated or the user can intentionally end the repeating step.

Claims (6)

In the method of extending ontology information,
Receiving, by the system, user-selected words included in the same topic;
Receiving, by the system, layer information and connection information (user-ontology information) between the user-selected words;
A system generating a syntax tree from a document included in the topic using the user-ontology information;
A system using the syntax tree information, determining valid-words, and generating word vector information; And
And a system using the word vector information to determine valid-similar words with high similarity to the user-ontology information, and combining them with the user-ontology information. According to claim 1,
To determine the effective-word of the step of determining the effective-word using the syntax tree information,
The system further comprises receiving user-ontology information and the effective-word distance information. According to claim 1,
To determine the effective-similar word,
And the system further comprising receiving variance values and deviation information between the variance values. In the method of extending ontology information,
Receiving, by the system, user-selected words included in the same topic;
Receiving, by the system, layer information and connection information (user-ontology information) between the user-selected words;
A system generating a syntax tree from a document included in the topic using the user-ontology information;
A system using the syntax tree information, determining a valid-word, and generating a selection-SVD-subgroup; And
A method of extending ontology information, comprising: a system combining the selected-SVD-subgroup and the user-ontology information. The method of claim 4,
In order to determine the effective-word in the step of determining the effective-word using the syntax tree information,
The system further comprises receiving user-ontology information and the effective-word distance information. The method of claim 4,
For the screening-SVD-subgroup determination,
The system further comprises the step of receiving the number of screening-SVD-subgroup ontology information extension method.

Images